2005-03-06 09:58:16

by jarmo

[permalink] [raw]
Subject: ax25 t1_timeout

Hi
Withs kernel 2.6.11 ax25 t1_timeout is working badly...
Expl. I have set t1_timeout to 10s (10000).Now testing..
Taking radio of,set connection to somewhere,first try then
10s time second try...But then 20s and try next 30s and try...
So t1_timeout is increasing?In 2.6.10 timeout is 10 and
it does not increase...Possible bug?

Jarmo
--
Linux is the answer.
What was your question?
(Brian Lane)


2005-03-07 15:16:34

by Ralf Baechle

[permalink] [raw]
Subject: Re: ax25 t1_timeout

On Sun, Mar 06, 2005 at 11:58:11AM +0200, jarmo wrote:

> Withs kernel 2.6.11 ax25 t1_timeout is working badly...
> Expl. I have set t1_timeout to 10s (10000).Now testing..
> Taking radio of,set connection to somewhere,first try then
> 10s time second try...But then 20s and try next 30s and try...
> So t1_timeout is increasing?In 2.6.10 timeout is 10 and
> it does not increase...Possible bug?

No, pilot error ;-)

Linux supports three different backoff types which can be configured via
/proc or sysctl. This is how to display the backoff type of interface
sp0, for example:

[root@dl5rb] sysctl net.ax25.sp0.backoff_type
net.ax25.sp0.backoff_type = 1
[root@dl5rb]

1 in this example stands for linear backoff which is the default and what
you've been observing. 0 would be no backoff which generally in a shared
RF environment is probably unfriendly and can mathematically be shown to
cause network meltdowns. Finally 2 would means exponential backoff,
which would use t1, 2*t1, 4*t1, 8*t1 for the waiting time, but never goes
worse than 8 * t1.

You can change the value with sysctl also but generally linear backoff is
quite reasonable.

73 de DL5RB op Ralf

--
Loc. JN47BS / CQ 14 / ITU 28 / DOK A21